- prompt from command line directly to Claude
- suggestions dumped into a file under ./tmp/ (ignored by git)
- iterate on those files
- shuttle test results over to Claude
Getting those files merged with the source files is also important, but I’m not confident in a better way than copy-pasting at this point.
I plan to share it on Github, but waiting for my employer's internal review process to let me open source it as my own project, since they can legally claim all IP developed by me while employed there.
Mainly - tool calling support just merged in llama.cpp (https://github.com/ggerganov/llama.cpp/pull/9639) this week, and it's been a fun exercise to put local LLMs through the wringer to see how they do at handling it.
It's been a mixture of "surprisingly well" and "really badly".
Save tens of hours in one commit?! No model is this good yet[1] - especially not Aider with its recommended models. I fully agree with parent - current SoTA models require lots of handholding in the domains I care about, and the AI chat/pairing setup works much better compared to the AI creating entire commits/PRs before a human gets to look at it.
1. If they were, Zuckerberg would have already announced another round of layoffs.
I often get hung up on UI, I can’t make a decision on what I think will look decent and so I just sort of lock up. Aider lets me focus on the logic and then have the LLM spit out the UI. Since I’ve given up on projects before due to the UI aspect (lose interest because I don’t feel like I’m making progress, or get overwhelmed by all the UI I’ll need to write) this is a huge boon to me.
I’m not incapable of writing UI, I’m just slower at it so Aider is like having a wiz junior developer who can crank out UI when I need it. I’m even fine to rewrite every line of the UI by hand before “shipping”, the LLM just helps me not get stuck on what it should look like. It lets me focus on the feature.
Also you can use local models if you want it to be “free”.
The thing is, all side projects have a cost, even if it’s just time. I’m happy to let some of my side projects move forward faster if it means paying a small amount of money.
I find Claude wants to edit files that I don’t like to be edited often. Two ways I deal with that - first, you can import ‘read only’ files, which is super helpful for focusing. Second, you can use a chat mode first to talk over plans, and when you’re happy say “go”.
I think the thing to do is try and use it at a fairly high level, then drop down and critique. Essentially use it as a team of impressive motivation and mid quality.
I do like the idea of letting the model ask for source code.
It’s all about attention / context.
I’m giving myself the option to output collated code to a file, or copy it to clipboard, or just hold onto it for the next prompt.
I know aider does this stuff, but because I’m automating my own workflow, it’s worth doing it myself.